Recognition of a subregion of human proinsulin by class I-restricted T cells in type 1 diabetic patients
- Andréa Toma*,
- Samy Haddouk*,
- Jean-Paul Briand†,
- Luc Camoin‡,
- Hanne Gahery‡,
- Francine Connan‡,
- Danielle Dubois-Laforgue*,§,
- Sophie Caillat-Zucman*,
- Jean-Gérard Guillet‡,
- Jean-Claude Carel*,
- Sylviane Muller†,
- Jeannine Choppin‡, and
- Christian Boitard*,§,¶
- *Institut National de la Santé et de la Recherche Médicale U561, Hôpital Cochin-Saint Vincent de Paul, Université Paris V, 75014 Paris, France; †Centre National de la Recherche Scientifique Unité Propre de Recherche 9021, Institut de Biologie Moléculaire et Cellulaire, 67000 Strasbourg, France; ‡Institut National de la Santé et de la Recherche Médicale U567, Centre National de la Recherche Scientifique Unité Mixte de Recherche 8104, Institut Cochin, Université Paris V, 75014 Paris, France; and §Service d'Immunologie Clinique, Hôpital Cochin-Saint Vincent de Paul, 75014 Paris, France
-
Communicated by Hugh O. McDevitt, Stanford University School of Medicine, Stanford, CA, May 24, 2005 (received for review March 15, 2005)
Abstract
Proinsulin is a key autoantigen in type 1 diabetes. Evidence in the mouse has underscored the importance of the insulin B chain region in autoimmunity to pancreatic beta cells. In man, a majority of proteasome cleavage sites are predicted by proteasome cleavage algorithms within this region. To study CD8+ T cell responses to the insulin B chain and adjacent C peptide, we selected 8- to 11-mer peptides according to proteasome cleavage patterns obtained by digestion of two peptides covering proinsulin residues 28 to 64. We studied their binding to purified HLA class I molecules and their recognition by T cells from diabetic patients. Peripheral blood mononuclear cells from 17 of 19 recent-onset and 12 of 13 long-standing type 1 diabetic patients produced IFN-γ in response to proinsulin peptides as shown by using an ELISPOT assay. In most patients, the response was against several class I-restricted peptides. Nine peptides were recognized within the proinsulin region covering residues 34 to 61. Four yielded a high frequency of recognition in HLA-A1 and -B8 patients. Three peptides located in the proinsulin region 41–51 were shown to bind several HLA molecules and to be recognized in a high percentage of diabetic patients.
Type 1 diabetes is characterized by the activation of lymphocytes against autoantigens expressed by pancreatic beta cells. T lymphocytes play a key role in the disease process. Diabetes has been reported in a patient deprived of B lymphocytes (1). In the nonobese diabetic (NOD) mouse, CD8+ T cells play a pivotal role in the initiation of autoimmunity (2). Beta 2 microglobulin-deficient NOD mice do not develop insulitis unless beta cell class I expression is restored (3). CD8+ T cells are responsible for beta cell destruction in transgenic mice over expressing beta cell-specific CD8+ T cells (2). In man, CD8+ T cells and IFN-γ-positive cells are major components of insulitis (4–9). Recurrent diabetes in recipients of isografts from a discordant identical twin is accompanied by predominant CD8+ T cell infiltration (10). However, few studies have characterized recognition of beta cell antigens by CD8+ T cells (11, 12).
Proinsulin has been ascribed a key role in diabetes. Insulin and proinsulin are targets of autoantibodies (13–15) and T cells (16–23) in diabetic and prediabetic subjects. Anti-insulin antibodies are the first autoantibodies detected in children at risk for diabetes (15). In the NOD mouse, transfer of insulin-specific T cells accelerates diabetes (24), and insulin exposure prevents diabetes (25). Diabetes development is altered in mice lacking the expression of proinsulin genes (26–28). Evidence in the mouse has underscored the importance of the proinsulin region encompassing the insulin B chain in diabetes autoimmunity (25, 29). CD8+ T cells that transfer diabetes in the NOD recognize an insulin B chain epitope (30). In man, a majority of proteasome cleavage sites are predicted within the insulin B chain and adjacent C peptide region, between residues 38 and 52, by proteasome cleavage algorithms (www.mpiib-berlin.mpg.de/MAPPP/cleavage.html), pointing to candidate peptides carrying correct COOH-termini within this region (31).
To characterize proinsulin peptides recognized by class I-restricted T cells, we selected 8- to 11-mer peptides by combining the characterization of proteasome cleavage patterns of the insulin B chain and adjacent C peptide region of proinsulin and the study of peptide binding to common class I molecules. We used these peptides to evaluate their recognition by peripheral blood mononuclear cells (PBMCs) from control and diabetic subjects.
Methods
Patients. Patients had type 1 diabetes (32) and islet cell antibodies, anti-GAD, anti-insulin, or anti-IA2 antibodies at diagnosis. Recent-onset patients (n = 19, male/female ratio 7/12, age 40 ± 19 years) were studied within 3 months of diagnosis. Long-standing diabetic patients (n = 13, male/female ratio 8/5, age 40 ± 5, duration of diabetes 23 ± 9 years) had been treated with insulin for >5 years at the time of study. Controls were 12 normal blood donors (male/female ratio 8/4) and 7 type 2 diabetic patients (male/female ratio 4/3), 3 of whom were treated with insulin (Table 1). An informed consent was obtained from all patients. PBMCs were isolated by Ficoll Paque density gradient centrifugation (Amersham Pharmacia Biotech) and analyzed within hours of sampling. HLA class I and class II alleles were determined by serological typing and genotyping, respectively.
Peptides. Peptides of 8 to 11 amino acids containing putative binding motifs for HLA-A1, -A2, -A3, -A11, -A24, -B8, and -B18 molecules were selected from the human proinsulin sequence, synthesized by using Fmoc chemistry, purified by RP-HPLC (33) and controlled by MALDI-TOF on a Protein TOF mass spectrometer (Bruker, Wissembourg, France). Proinsulin peptide nomenclature refers to N- and C-terminal positions along the human preproinsulin sequence.
HLA Class I Peptide-Binding Assay. HLA molecules were purified from Epstein–Barr virus (EBV)-transformed B cell lines (34, 35). Aliquots of HLA H (heavy) chains were incubated with 10-4,10-6, and 10-8 M exogenous peptide. Reassembled HLA molecules were incubated with anti-HLA monoclonal antibodies (BB7.2 for HLA-A2, GAP A3 and A11.1M for HLA-A3 and -A11, B1.23.2 for HLA-B molecules, and PA2.6 for HLA-A and -B molecules) and correctly folded HLA complexes were revealed (35). Percentage of binding was defined as binding of tested peptide over binding of reference viral peptide × 100. Reference peptides were influenzae virus matrix M.58–66 (GILGFVFTL) for HLA-A2, HIV Nef 73–82 (QVPLRPMTYK) for HLA-A3 and -A11, influenzae virus PB1 peptide 591–599 (VSDGGPNLY) for HLA-A1, EBV LMP2 peptide 419–427 (TYGPVFMCL) for HLA-A24, human papilloma virus E2 peptide 44–52 (QAEPDRAHY) for HLA-B18, and HIV Nef 90–97 (FLKEKGGL) for HLA-B8.
Enzyme-Linked Immunospot (ELISPOT) Assay. The IFN-γ ELISPOT assay was performed as described in ref. 36. PBMCs were suspended in complete medium supplemented with 10% FCS (PAN Biotech, Aidenbach, Germany), plated in triplicate at 3 × 105 cells per well and incubated overnight in the presence of 10 μg/ml peptide and 25 units/ml human IL-2. Spots were counted by using a KS ELISPOT device (Zeiss, Hallbergoos, Germany).
Background IFN-γ response was evaluated in 3–6 wells (3 × 105 cells per well) in the absence of a peptide. Responses were considered positive when the number of spot-forming cells (SFC) in the presence of peptide was above background plus 3 SD. Positive controls consisted of three wells (3 × 104 cells per well) stimulated with 1 μg/ml phytohemagglutinin and three wells (3 × 105 cells per well) stimulated with 10 μg/ml of reference viral peptides. Negative controls were HIV Nef peptides (Table 2). To compare responses of PMBCs to allele/peptide complexes, a stimulation score (SS) was calculated to take into account interassay variability (SS = mean number of spots in response to peptide - mean number of spots in absence of peptide). Comparison of SSs of patients and controls used nonparametric Mann–Whitney test.
Proteasome Digestion, Separation of Peptides, and Mass Spectrometry Analysis. Large proinsulin peptides covering the B chain and adjacent C peptide (peptide 28–57 and 32–64, Fig. 1a) were digested by a proteasome-enriched extract obtained from T1 lymphoblastoid cells. Cleavage products were separated by RP-HPLC (PerkinElmer, Norwalk, CT). Chromatograms were recorded at 214 nm (36). Fractions containing cleavage products corresponding to HPLC peaks and regions between peaks were lyophilized before mass spectrometry analysis or functional tests.
Response of two patients to proinsulin peptides 34–42 and 41–50. (a) Proinsulin region 28–64 and predicted proteasome cleavage sites. (b and c) IFN-γ ELISPOT response of patients R1 and L1, respectively. Columns are mean ± SD of triplicate wells.
Mass analyses were performed on a MALDI-TOF spectrometer in a reflectron delayed extraction ion source over a mass range of 500–3,200 Da and recorded with a Voyager-DE-Pro mass spectrometer (PerSeptive Biosystems, Framingham, MA). Monoisotopic masses were calculated after calibration with three external standards (des-Arg-1-Bradykinin, [M + H]+ 904.4681; angiotensin I, [M + H]+ 1296.6853; and Glu-1-Fibrinopeptide B, [M + H]+ 1570.6774) (PerSeptive Biosystems). Peptides corresponding to computed masses were identified taking into account methionine oxidation, with gpmaw 4.2 software (Lighthouse data, Odense, Denmark), with ±50 ppm mass accuracy.
To detect antigenic peptides in proteasome digests, thawed PBMCs from two patients were tested by using IFN-γ ELISPOT assay. Lyophilized fractions were dissolved in 400 μl of complete medium and 100 μl per well were added in each well.
Results
Detection of T Cells Specific for Proinsulin 34–42 and 41–50. Peptides presented by MHC class I molecules are processed by the intracellular proteolytic machinery. The proteasome is important for COOH-terminal cleavage of cytosolic peptides. Our study focused on proinsulin region 28–64 (Fig. 1a). To set up an assay to test the recognition of proinsulin peptides by PBMCs from patients, the proinsulin 28–64 sequence was analyzed for the presence of 8- to 11-mer peptides carrying C-terminal residues predicted by proteasome cleavage algorithms (http://mpiib-berlin.mpg.de/MAPPP/cleavage.html) and binding motifs for two common class I molecules, HLA-A2 and -A1. Peptides 34–42 and 41–50 were selected and tested for recognition by PBMCs from patients R1 (HLA-A2-B8) and L1 (HLA-A1-B18) (Table 1) by using an IFN-γ ELISPOT assay. Patients R1 and L1 (R, recent-onset diabetes; L, long-standing diabetes) showed an IFN-γ-positive response to peptides 34–42 and 41–50, respectively (Fig. 1 b and c). Patient R1 also responded to M.58–66 but not to Nef 83–91. Patient L1 responded neither to NP 44–52 nor to Nef 121–128. To define responder cells, T cells were purified from patient L1's PBMCs. In the presence of peptide 41–50, a positive response of CD4+-depleted PBMCs was observed, although not of CD8+-depleted PBMCs, suggesting that CD8+ T cells were responsible for the IFN-γ production (data not shown).
Proteasome Processing of Proinsulin Peptides. We analyzed peptides resulting from proteasome digestion of two peptides covering proinsulin region 28–64. A first peptide (28–57) was incubated with a proteasome-enriched T1 cell extract at 37°C for 4 and 20 h (36). Several digestions yielded identical HPLC profiles. A 20-h incubation was retained for collecting fractions. Multiple peaks appeared between 20 and 70 min and a late peak, corresponding to undigested peptide 28–57, was evidenced. Fractions collected between 29 and 63 min (fractions 1 to 8, Fig. 2a) were tested in the presence of PBMCs from patients R1 and L1. IFN-γ positive responses were observed to HPLC fractions 1, 3, and 4 in patient R1 (Fig. 2b) and to fractions 3 and 4 in patient L1 (Fig. 2c). No response was observed to peptide 28–57, as expected for CD8+ T cell responses.
Epitopes defined in proteasome digests of proinsulin 28–57. (a) HPLC profile of peptide 28–57 digests by proteasome-enriched T1 cell extract. (b and c) ELISPOT responses patient R1 and L1 PM-BCs, respectively, to HPLC fractions. The left column is incubation in the absence of peptide. Each fraction is given an ordered number (1–8) indicated in a and under corresponding columns in b and c. In the last three columns, incubation in the presence of (i) undigested peptide 28–57, (ii) influenza virus peptide M.58–66 or NP 44–62, and (iii) phytohemagglutinin (PHA), respectively. Columns are mean ± SD of duplicate wells. The dotted line represents the background response plus 3 SD.
Mass spectrometry analysis of HPLC fractions 1, 3, and 4 characterized 8- to 15-mer peptides carrying C-termini that correspond to cleavage after Leu-39, Phe-48, Tyr-50, Thr-51, Thr-54, Arg-55, and Arg-56 (Table 3). PBMCs from patient R1, but not L1, responded to fraction 1, which contained a single peptide, proinsulin 49–57. This peptide had several basic residues but no adequate C-terminal anchoring residue for HLA-B8. PBMCs from patient L1 mainly recognized fraction 4, which contained several peptides resulting from cleavage after Thr-51.
These data confirmed the predicted sensitivity to proteasome of the C-terminal half of the B chain and adjacent C peptide. Given recognition of proinsulin 49–57 in fraction 1 by patient R1, fractions resulting from proteasome digestion of peptide 32–64 were analyzed. Total digests of proinsulin peptides 28–57 and 32–64 confirmed the numerous cleavage sites located in 39–57 and the cleavage after Glu-57. New cleavage sites after residues Val-42, Arg-46, Lys-53, Ala-58 and Leu-61 were evidenced. Several cleavage sites that were not predicted by algorithms were characterized, i.e., Leu-39, Lys-53, Thr-54, Arg-55, Arg-56, Glu-57, and Ala-58.
These results defined candidate epitopes that carry proteasome C-terminal cleavage residues and binding motifs for HLA-A1, -A2, -A3, -A11, -A24, -B8, and -B18 molecules (Tables 3 and 4).
Binding of Proinsulin Peptides to HLA Class I Alleles. Thirteen 8- to 11-mer peptides carrying consensus or related binding motifs to HLA-A1, -A2, -A3, -A11, -A24, -B8, and -B18 (37, 38) were tested for binding to purified HLA class I molecules (Table 5) along with reference viral peptides (data not shown). Nine peptides showed significant binding to at least one HLA class I molecule. Peptides that yielded >50%, 20–50%, and <20% binding at 10-6 M were considered high, intermediate, and low binders, respectively. Binding of all but one peptide (proinsulin 38–46) was weaker than that of control viral peptides. Four peptides (38–46, 41–50, 42–51, and 44–51) bound to several HLA-A and/or HLA-B alleles.
Recognition of Viral and Proinsulin Peptides by T Cells. Proinsulin peptides defined as binders to HLA class I molecules and control peptides were tested for recognition by PBMCs from control subjects and type 1 diabetic patients (Table 6).
Responses to Viral Peptides. We studied the response to known viral CD8+ T cell epitopes in our study population. An IFN-γ ELISPOT response to influenzae matrix peptide M.58–66 that ranged between 3 and 20 SD above background was observed in 8 of 11 HLA-A2 controls and 13 of 14 HLA-A2 patients. A response to nucleocapsid influenzae peptide NP 380–388 that ranged within 3 and 10 SD above background was observed in 2 of 7 HLA-B8 controls and 7 of 11 HLA-B8 patients. A response to nucleoprotein influenzae peptide NP 44–52 that ranged between 3 and 8 SD above background was observed in 3 of 3 HLA-A1 controls and 3 of 5 HLA-A1 patients. No response was observed to HIV peptides in HLA-A1, -A2, and -B8 individuals (response <2 SD) (Table 2).
Responses to Proinsulin Peptides. Given the detection of responses to viral peptides and our preliminary results of responses to proinsulin peptides 34–42 and 41–50 in patients R1 and L1, we evaluated the response to proinsulin peptides in diabetic patients. A positive IFN-γ response was observed against at least one proinsulin peptide in 29 of 32 (17 of 19 recent-onset and 12 of 13 long-standing) type 1 diabetic patients. The prevalence of responses to proinsulin peptides was comparable in recent-onset and long-standing diabetic patients. No response was observed in control individuals. When pooling tests performed in controls against all 16 peptide/allele complexes indicated in Table 6, 51 of 55 and 4 of 55 showed a number of spots in the presence of peptide that ranged from 0.0 to 1.0 and 1.0 to 1.86 SD, respectively, above background. When pooling patient tests, 59.41% showed a number of spots in the presence of peptide that was 3 SD above background. The percentage of positive tests was 60.45%, 53.2%, 48.4%, and 40.4% when setting the threshold at 2, 4, 5, and 6 SD above background, respectively. When considering peptides for which at least 5 patients carrying a defined class I allele were tested, five peptides were recognized in >50% of patients: proinsulin 41–50 (HLA-A1 and -A3), 42–51 (HLA-A1, -A2, -B8, and -B18), 44–51 (HLA-A1 and -B8), 45–53 (HLA-A3), and 49–57 (HLA-B8). Proinsulin 34–42 was recognized by 43% of patients. By contrast, three proinsulin peptides (38–46, 39–48, and 51–61) were recognized in <30% of the patients (Table 6). When considering allele/peptide complexes for which at least four patients and four control individuals were tested, the mean number of spots in patients was significantly higher that that in controls in response to peptides 34–42 (HLA-A2), 41–50 (HLA-A1), 42–51 (HLA-A2 and -B8), 44–51 (HLA-B8), and 49–57 (HLA-B8) (Table 6).
Proinsulin peptides 42–51 and 44–51 were recognized in >70% of tested patients, probably in association with several HLA molecules (Table 6). A high prevalence of positive responses (>50% of patients) was observed with 3 of 5, 6 of 8, and 2 of 3 peptides yielding high, intermediate, and low HLA binding, respectively. A moderate response prevalence (25–50%) was observed, with 1 of 5 and 2 of 8 peptides yielding high and intermediate HLA binding, respectively. A low response prevalence (<25%) was observed with 1 of 5 and 1 of 3 peptides yielding high and low HLA binding, respectively. There was no correlation between the response prevalence to proinsulin peptides and the strength of binding to corresponding class I alleles. All 12 nondiabetic controls and 7 type 2 diabetic patients, of whom 3 were treated with insulin, showed negative responses (<1.86 SD) to the 9 proinsulin peptides.
PBMCs from 14 diabetic patients were tested 2–4 times within a 3-month time frame. A high reproducibility of positive responses (100% positive tests) was observed in 10 of 14 patients who showed a positive IFN-γ response to at least one proinsulin peptide (38–46, 42–51, 44–51, and 51–61) in the first test performed. Interestingly, the response to peptide 51–61 was highly reproducible (4 of 4 positive tests) in one HLA-B8 patient (R1) (data not shown), although 11 other HLA-B8 patients were nonresponders. Overall, the reproducibility of positive responses to proinsulin peptides was comparable with that against viral peptides.
Discussion
Defining beta cell antigen epitopes is expected to have wide-range implications in the development of T cell assays and the design of peptide-based immunotherapy in type 1 diabetes. Although numerous epitopes presented by HLA class II alleles to CD4+ T cells have been defined in the human (20) and in transgenic mice expressing human class II alleles (39–41), only two epitopes presented by class I alleles to CD8+ T cells have been defined, an HLA-A2-restricted GAD epitope (11) and an HLA-A2 epitope (IAPP5–13) derived from islet amyloid polypeptide (12). Using a systematic approach to define proinsulin class I epitopes, we bring evidence that PBMCs from diabetic patients secrete IFN-γ in response to 8- to 11-mer proinsulin peptides. We focused our study on a region of proinsulin in which proteasome-mediated cleavage sites and anchoring motifs for class I molecules that are common in type 1 diabetic patients were predicted. Peptides were selected as carrying proteasome C-terminal cleavage sites and as binding to either HLA-A1, -A2, -A3, -A11, -A24, -B8, or -B18 class I molecules. Some peptides bound to several class I molecules, as proinsulin 38–46 for HLA-A3 and -A11; 41–50 for HLA-A1, -A3, and -A11; 42–51 for HLA-A1, -A2, -B8, and -B18; and 44–51 for HLA-A1 and -B8. All selected epitopes within proinsulin region 28–64 were recognized by PBMCs from diabetic patients. The prevalence of recognition of at least one proinsulin peptide was >90%. Four peptides (41–50, 42–51, 44–51, and 49–57), three of which located in region 41–51, were recognized by a high percentage of HLA-A1 and -A3; HLA-A1, -A2, -B8, and -B18; HLA-A1 and -B8; and HLA-B8 diabetic patients, respectively. The absence of recognition of proinsulin peptides by PBMCs from control individuals or type 2 diabetes patients makes a strong case for the possible role of proinsulin-specific CD8+ T cells in type 1 diabetes. Two epitopes that were recognized by PMBCs from diabetic patients were located within a region overlapping the B chain and C peptide and within the C peptide (49–57 and 51–61, respectively), indicating recognition of proinsulin rather than exogenous insulin. None of insulin-treated type 2 diabetic controls showed PMBC recognition of insulin peptides.
In most patients, the response to proinsulin was multiepitopic. However, the long preclinical phase that precedes clinical diabetes does not preclude that a more restricted set of peptides is recognized at initiation of the autoimmune process (29). More surprisingly, proinsulin peptides were recognized both in recent-onset and long-standing diabetic patients. This finding may reconcile with previous observations in type 1 diabetes in which a dramatic recurrence of diabetes is seen in recipients of a hemipancreas graft from monozygotic diabetes-discordant twins with an almost exclusive CD8+ T cell infiltrate (10). This observation, along with our data, may indicate that long-term memory class I-restricted T cells persist in patients who have been deprived of residual beta cells for years. In the NOD mouse in which no T cell activation is seen in mice deprived of beta cells at an early stage of the autoimmune process, full activation of T cells is seen in the absence of residual beta cells once the autoimmune process is initiated (42).
Using a strategy based on peptide library-mediated in vitro assembly of class I molecules, proinsulin peptides have already been defined on the basis of their association with HLA-B8, -A2, and -B15. Several candidate epitopes were shown either to harbor anchor residues that were only weakly predicted or not predicted by commonly used algorithms or not to contain canonical allele-specific binding motifs (43). In our study, some of these peptides were recognized by PBMCs from >40% of patients, such as proinsulin 44–51 in HLA-B8 patients or 34–42 and 42–51 in HLA-A2 patients.
Intracellular mechanisms that define the amino acid sequence of peptides presented to CD8+ T cells include proteolytic breakdown of intracytosolic peptides by the proteasome, translocation by transporter-associated with antigen processing and binding of peptides into the groove of MHC class I molecules within the endoplasmic reticulum (31). The strategy that we applied was based on the characterization of proteasome cleavage sites that generate the appropriate C-termini of potential T cell epitopes, as previously reported in case of viral epitopes (44, 45). We used a 20S proteasome-enriched preparation extracted from the T1 lymphoblastoid cell that contains IFN-γ-inducible LMP-2 and LMP-7 proteasome subunits. Some peptides characterized in proteasome digests corresponded to epitopes of optimal size that could directly bind to class I molecules, as possibly proinsulin 30–39, 34–42, 44–51, and 49–57. The generation of a majority of peptides as N-terminally extended precursors of epitopes was expected, given the possibility of further trimming of precursors by aminopeptidases (46). Peptides were selected by combining analysis of proteasome cleavage and of binding to purified class I molecules in vitro. Among 10 major proteasome cleavage sites that we defined, 6 were predicted by using a proteasome cleavage prediction algorithm. We did not take into account the frequency of cleavage sites generated by proteasome digestion, because no relation with generation of epitopes derived from self-antigens has been established. Seemingly, we did not restrict our study to epitopes showing the highest binding affinity to purified class I molecules because the rules governing the nature of epitopes presented in autoimmunity remain far from clear. There was no clear correlation between the prevalence of responses to proinsulin peptides and the affinity levels of peptide binding to purified HLA class I molecules. Frequent responses to low-affinity peptides were observed as proinsulin peptides 44–51 in HLA-B8 individuals and 42–51 in HLA-A2 patients.
Our data bring evidence that a majority of type 1 diabetic patients shows IFN-γ responses to 8- to 11-mer peptides that are derived from proinsulin, a major beta cell autoantigen. We also bring evidence that T cell responses are detected in long-term diabetic patients. Beyond their capacity to produce IFN-γ, effector functions of proinsulin-specific T cells and their possible participation to beta cell destruction will have to be delineated.
Acknowledgments
This work was supported by a recurrent grant from Institut National de la Santé et de la Recherche Médicale and from Ministère de la Recherche et de la Technologie Grant 03 246.
Footnotes
-
↵ ¶ To whom correspondence should be addressed. E-mail: boitard{at}paris5.inserm.fr.
-
Abbreviations: PBMCs, peripheral blood mononuclear cells; NOD, nonobese diabetic.
- Copyright © 2005, The National Academy of Sciences
